SplitFed: When Federated Learning Meets Split Learning

Releasing the source code version1 of our work "SplitFed: When Federated Learning Meets Split Learning."

We have three versions of our programs:

Version1: without using socket and no DP+PixelDP

Version2: with using socket but no DP+PixelDP

Version3: without using socket but with DP+PixelDP (required more packages)

Other versions will be released soon.

Please cite the main paper if useful: https://arxiv.org/pdf/2004.12088.pdf

Description

This repository contains the implementation of Centralized Learning (baseline), Federated Learning, Split Learning, SplitFedV1 Learning and SplitFedV2 Learning.

All programs are written in python 3.7.2 using the PyTorch library (PyTorch 1.2.0).

Dataset: HAM10000

Model: ResNet18